1 An Initial Reconstruction of Proto-Lampungic: Phonology and Basic Vocabulary 1 Karl Anderbeck SIL International karl_anderbeck<at>sil<dot>org Abstract Lampungic isolects such as Komering, Pubian and Lampung are spoken by approximately one and a half million people in the western Indonesian provinces of Lampung and South Sumatra. To date there has not been any published reconstruction of earlier stages of Lampungic nor a detailed investigation of its classificatory status within Western Malayo-Polynesian (WMP). Differing subgrouping hypotheses for Lampungic have been posited, whether with Malayic, or Malayo-Javanic, or, most recently, as an isolate. A major problem has simply been the paucity of published data. Drawing on a recent SIL dialect study as well as earlier sources, this study is an initial attempt to provide parameters for understanding the place of Lampungic within WMP. Specifically, it is a reconstruction of the phonology and basic vocabulary of what is called Proto-Lampungic (PLP). Drawing on the disciplines of dialectology, comparative linguistics and previous reconstructions of relevant proto-languages, a principled distinction between pre- and post-plp innovations is made. The result is a bundle of features which together can be used to define a Lampungic subgroup and to distinguish it from its neighbors. A subgrouping hypothesis is advanced which groups Lampung and Sundanese together on the basis of shared phonological innovations. Beyond a discussion of its place in WMP, a brief treatment is given of Lampungic s contribution to the discussion of disputed PMP segments such as *r and *-ey. 1 Many people have contributed in one way or another to the development of this paper. I wish to thank the Indonesian Institute of Sciences for providing visas and permissions for this research, SIL International for providing the funding, my SIL colleagues who collected much of the data used herein, Chad White who wrote much of the first draft of this paper, and the Lampungic-speaking people who were so generous with their time and help, even protecting us from being robbed of our survey equipment! I would like to single out my friend Mas Ali from Menggala, Lampung, as someone who went above and beyond to help us dig deeper into the secrets of his language.
2 2 Contents Abstract... 1 Contents... 2 List of Tables... 4 List of Maps... 5 List of Abbreviations Introduction External classification History and ecology of Lampung language Internal classification Previous research Dialectology Comparative method WordCorr Lampung today Data sources Phonology Consonants Vowels Dialect differences in detail Realization of *ə Vowel lowering Nasal consonant cluster reduction Deletion of initial *h Fortition of final *h Final diphthongs Survey findings Reconstruction of PLP phonemes and word structure PLP consonant phonemes *b (Labial Plosives) *p (Labial Plosives) *d (Laminal Plosives) *t (Apical Plosives) *g (Dorsal Plosives) *k (Dorsal Plosives) *ʔ (Postdorsal Plosives) *m (Labial Nasals) *n (Laminal Nasals) *ɲ (Laminal Nasals) *ŋ (Dorsal Nasals) *l (Laminal Resonants) *c (Laminal Fricatives) *j (Laminal Fricatives) *s (Laminal Fricatives) *r (Dorsal Fricatives) *h (Postdorsal Fricatives)...39
3 *y (Laminal Approximants) *w (Dorsal Resonants) PLP vowel phonemes *i (Close Front) *u (Close Back) *ə (Open-Mid Central) *a (Open Central) *uy (Close Back) *ay (Close Front) *aw (Open Back) PLP word structure and phonotactic constraints PLP lexicon PLP lexical reconstructions Pronouns and demonstratives Identical forms Initial nasals and word classes PLP *ay and *aw reconstructions Discussion of reconstructions PLP reduplicated stems Loanwords Changes from PMP to PLP Innovations and retentions in consonant phonemes PMP *q to PLP *h PMP *h PMP and PHN *ʔ PMP *w Nasal excrescence Medial consonant clusters PMP *R to PLP *i PMP *d > PLP *r PMP *j PMP *z to PLP *j Innovations and retentions in vowel phonemes PMP *uy PMP *-ay/*-ey PMP *-aw /*-ew PMP *-iw Low-high vowel sequences Rule ordering Lexical phenomena PLP s place in Malayo-Polynesian Relationship with Batak Relationship with Malayic Relationship with Malayo-Chamic-Bali-Sasak-Sumbawa Relationship with Sundanese Conclusions
4 4 6.1 Results Linguistic diversity and potential time depth vis-à-vis Batak and ML Suggestions for further research References List of Tables Table 1 Wordlist sites Table 2 Basic consonant phonemes of the Lampungic cluster Table 3 Examples of debuccalization Table 4 Basic vowel and diphthong phonemes of the Lampungic cluster Table 5 Examples of ultimate *ə > [o] Table 6 Examples of split ultimate *ə realization in Nyo areas Table 7 Examples of ultimate *ə in non-oral environments in Blambangan Pagar (KotaBumi). 17 Table 8 Examples of penultimate *ə > [o] Table 9 Examples of *NS > N with voiced stops Table 10 *h deletion word initial Table 11 *h deletion word final Table 12 Final diphthongs Table 13 Final diphthongs and bound morphemes in Menggala Table 14 Consonant phonemes of PLP Table 15 Lampung WordCorr consonant environments Table 16 Lampung WordCorr vowel environments Table 17 Distribution of PLP phonemes Table 18 Disallowed medial vowel sequences Table 19 Reconstructed PLP wordlist Table 20 PLP pronouns and demonstratives Table 21 PLP many and if Table 22 PLP reduplicated stems Table 23 Loanwords in LP Table 24 LP words with alveolar r Table 25 Word-initial PMP *q > PLP *h Table 26 Word-medial and word-final PMP *q > PLP *h Table 27 Reflexes of PMP *w Table 28 Nasal/liquid excrescence from PMP Table 29 Nasal consonant cluster reduction Table 30 Word medial PMP *R changes Table 31 Word medial PMP *er changes Table 32 Word final PMP *R changes Table 33 Tabulation of PMP *R reflexes Table 34 PMP initial *R reflexes Table 35 PMP *r words Table 36 Change of PMP *d to PLP *r Table 37 Medial PMP *j reflexes Table 38 Final PMP *j reflexes Table 39 PLP reflexes of PMP *z and *Z
5 5 Table 40 Retention of PMP *-uy Table 41 Final ǝy diphthongs in Nyo Table 42 PLP *-ay/*-ey reflexes Table 43 LP *-ay reflexes Table 44 Retention of PMP *-aw List of Maps Map 1 Major dialect groupings within LP...9 Map 2 LP wordlist sites...13 Map 3 Penultimate and ultimate *ə...19 Map 4 Initial *h loss and final *h fortition...21 Map 5 Final diphthongs and final *a...23
6 6 List of Abbreviations Language varieties most frequently referred to in this monograph. Primary sources are listed; when other sources are used this is noted in the text. AN - Austronesian AR - Arabic (Jones 1978) JV - Javanese (Horne 1974) LP - Lampung(ic) language ML - Malay (in general and including Sumatran Malay; various sources) PHF - Proto-Hesperonesian-Formosan (Zorc 1995) PHN - Proto-Hesperonesian (Zorc 1995) PLP - Proto-Lampungic PM - Proto-Malayic (Adelaar 1992) PMP - Proto-Malayo-Polynesian (primarily Blust 1999 and Zorc 1995; also Blust , Adelaar 1992) SI - Standard Indonesian (Echols & Shadily 1989) SKT - Sanskrit (de Casparis 1997) SM - Standard Malay (Wilkinson 1959) Other abbreviations and symbols: adj - adjective C - consonant id. - identical IPA - International Phonetic Alphabet n - noun N - nasal n.d. - no date/unpublished PL - plural - section S - consonant stop (i.e. plosive) SG - singular sth. - something v - verb V - vowel
7 7 1 Introduction Lampung is a set of related Austronesian language varieties spoken by around one and a half million people on the southern part of the island of Sumatra. There are many unanswered questions about the history of the Lampung people and their language; how long has this group been where it is? Within their language, is there any evidence of migrations, and if so, from where? What relationship does Lampung have with other Austronesian languages including its neighbors? This comparative study will certainly not answer all the questions above, but is aimed at providing initial parameters for understanding the history of the Lampung (LP) language or languages, namely a reconstruction of parts of the phonology and lexicon of Proto-Lampungic, how it developed from Proto-Malayo-Polynesian (PMP), and how it changed into the various forms we see currently. It is written to be complementary to the Lampung survey report (Hanawalt et al. in progress) but will still be intelligible on its own. 1.1 External classification Dyen s A LEXICOSTATISTIC CLASSIFICATION OF THE AUSTRONESIAN LANGUAGES (1965) was one of the first to tackle the classification of Lampungic (LP). Under his SUNDIC HESION he put the Javo-Sumatra Hesion, Sasak, Balinese, the Malayic Hesion, and the Dayak Subfamily among others. He originally grouped LP under the MALAYIC HESION along with Madurese, Achehnese, Minangkabau, and Kerinci (1965: 26). Since Dyen s research the term MALAYIC has been constricted to apply only to Malay dialects and their close kin like Iban (Adelaar 1992). Also, being that Dyen s classifications were based on lexicostatistics and given the large amount of Malay borrowings in LP, it has since become clear to scholars that LP is not as closely related to Malayic as would be implied by lexicostatistics. Malcolm Ross (1995:78) gives 24 groups for the Western-Malayo-Polynesian languages. He writes, Group 18 contains only Lampung, of extreme south-east Sumatra. Although it has been suggested in the past that it belongs to the Malayic group, current opinion regards it as not yet classified (Blust, pers. comm., Nothofer 1988). Adelaar (2005a) brings forward Ross s classification of LP without further comment. The question therefore remains relatively unexplored whether LP can be subgrouped with Malay or with any other language below the (Western) Malayo-Polynesian level. What I will attempt to demonstrate in this paper is that the defining characteristics of Lampungic are: loss of PMP *h in all positions PMP *q > PLP *h Proto-Hesperonesian *ʔ > PLP *h retention of PMP *w very limited medial nasal excrescence limited consonant cluster reduction merger of PMP *R and *r in word-initial position non-initial PMP *(e)r > PLP *y conditioned merger of PMP *j and *d with PLP *r retention of PMP *z as PLP *j retention of the PMP four-vowel system, and of diphthongs *-ay, *-aw and *-uy [irregular areal feature] shift in some instances of PMP *-ay and*-aw to i and u respectively PMP *-iw > PLP *(y)u epenthetic semivowel w or y inserted between low-high vowel combinations
8 8 Nothofer s (1985:294) SYSTEM 3 PAN numeral system The question will also be raised whether any of Adelaar s (2005a) 22 other WMP subgroups seems to share enough of these features to make a convincing case for merger with another subgroup or for positing a shared intermediate node under WMP. 1.2 History and ecology of Lampung language Comparative linguistics, archaeology and other disciplines have given us an understanding of the origins and general migration patterns of early Austronesian-speakers and approximately when they began to move into the regions we now call island Southeast Asia. It is thought that these speakers brought agricultural technology with them, which allowed them (or at least their languages) in many cases to overwhelm the earlier inhabitants of the areas they entered. Bellwood (1999) believes that the lowland areas of Sumatra were probably quite thinly populated due to their unsuitability for foraging prior to the advent of agriculture, so it may not have been too difficult for groups like Lampung to establish themselves in this new territory without enduring the type of language interference we see in, say, eastern Indonesia. However, history gives us a picture of substantial later contact between Lampung and other language groups including Javanese, Sanskrit, and, of course, Malay. In fact, for most of the past two millennia Lampung must have been under at least some level of domination by either the Malay port authorities to its north like Srivijaya, or the agrarian Javanese kingdoms to its south and east. I argue in this paper that it is Malay influence which is of the highest degree in Lampung vis-à-vis the other potential donor languages. Sumatran Malay-speaking groups border and partially surround Lampung. Additionally, in modern times we have Indonesian, the 800-pound linguistic gorilla. Walker writes in his 1976 grammar of Way Lima, Lampung, The influence of the Indonesian/Malay language on Lampung is pervasive. Contacts with Malay go back hundreds of years. In the past decades the influence of the national language is even stronger, affecting the phonology, the grammar, and the lexicon of Lampung. 4 (PLP lexicon) and 5 (Changes from PMP to PLP) will give substantial attention to the thorny issue of teasing out ML borrowings from what is truly Lampungic. 1.3 Internal classification On the basis of compared sound systems, lexicon, sociolinguistic attitudes, and reported and measured proficiency, Hanawalt et al. (in progress) conclude that LP can be divided into three major dialect clusters: 1. Lampung Api 2. Komering 3. Lampung Nyo LAMPUNG API and NYO are named after their respective words for what, while KOMERING is the name of the river which forms the homeland of the northernmost dialect cluster. LAMPUNG API (henceforth Api) is also often referred to as PESISIR, meaning coastal, while NYO is often referred to as ABUNG, which is an important ethnonym within the NYO grouping. See Map 1 for a visual illustration of their locations.
9 9 Map 1 Major dialect groupings within LP There are really not any strong linguistic differences between Api and Komering; their relationship is more of a language chain than two completely separate clusters. The starkest differences are between Nyo and the rest of LP. Dialect differences will be discussed in further detail in Previous research Hanawalt et al. (in progress) lay out in detail the various linguistic and sociolinguistic studies that have been done on Lampungic isolects, so that will not be repeated here. I will
10 10 just mention studies that have been of particular benefit to this historical-comparative look. There have been four dictionaries published in some form, one (Gaffar et al. n.d.) Komering to Indonesia, two others (Junaiyah 1985 and Hadikusuma 1994) focused on the Abung dialect group. Noeh et al. (1979) is the most comprehensive and provides information on a few different Api groups as well as Nyo. The others are not very comprehensive; Hadikusuma helpfully includes some information from 14 dialect areas but is the thinnest of the bunch. I relied on two phonological descriptions, one of a Lampung Api area called Way Lima (Walker 1976) and the other of Komering (Abdurrahman and Yallop 1979). Arguably the most helpful resource for this study was Walker s (1975) A LEXICAL STUDY OF LAMPUNG DIALECTS, which included 12 wordlists and an initial discussion of internal dialect divisions based on lexicostatistics. As part of a larger work on the people and agriculture of Sumatra (Tsubouchi 1980), Yasuyuki Mitani wrote on the language varieties of South Sumatra Province. In his paper he commented on Walker s classification of the Lampung dialects and offered a theory of the history of the Lampung dialects. Unfortunately he did not publish the data upon which his conclusions were based. Aliana et al. (1986) describe thirteen speech varieties within Lampung Province using lexicostatistics to determine the most central variety. Also informative was Hadikusuma s (1996) ADAT ISTIADAT DAERAH LAMPUNG ( Custom and Tradition in the Lampung Area ) which discusses some of the ethnic divisions and their backgrounds. Most of the data used in this paper were gathered by SIL personnel over the period Linguistic instruments included a 350-item wordlist, a sentence elicitation list, and targeted phonological and historical-comparative elicitation. 1.5 gives greater detail on the WHERE and WHY of data sources used in this paper. 1.5 Dialectology Dialectology is concerned with defining dialect boundaries and developments within the language family. Collins (1989:237) says it this way. The task of the dialectologist is to identify the splits which have yielded the contemporary network of dialects. In other words, delineating the history of a language, its diffusion, and its diversification, is the goal of dialectology. There are two basic models that have been followed to explain such divisions: the tree model and the wave model. The tree model assumes a sharp split has occurred by a separation or migration of the language community. The wave model sees innovations like pebbles dropping into a pond of water. The ripples that are created move ever outward creating a welter of isoglosses that crisscross one another within the area (Chambers and Trudgill 1998:91). This is sometimes referred to as diffusion. This comparative study is firmly set within a dialectological foundation. In a reconstruction of a language such as Lampung, where the entire language community lives within a single (albeit large) geographical area, one can expect to see substantial diffusion of linguistic innovations across dialect areas. The better the sampling, of both innovative and relic dialect areas, the further back in history one is able to go with a reconstruction, and the greater one s ability will be to accurately generalize to the whole language group. 1.6 Comparative method We will use the historical-comparative method to discover innovations within Lampung and retentions from PLP. In the comparative method we find regularly recurring correspondence sets that occur across cognate sets of the forms found in the speech varieties being studied. These correspondence sets give an indication of what the original sound in the proto-language was. Establishing a set of these sounds we can begin to form the lexicon and phonology of the proto-language. Working from the proto-language back to the present day forms of the
11 11 language we can see where changes have taken place (innovations) and where the segment has stayed the same (retentions). It is those retentions and innovations that help us determine the language classification and define dialect boundaries. What I am seeking to define, through the use of the comparative method, is a significant intermediate stage between Proto-Malayo-Polynesian and the present where Lampung had differentiated itself from its (Western) Malayo-Polynesian kin but not yet undergone significant internal dialectal differentiation. That this was ever the case is a big assumption, but we will see that it holds up rather well as a working hypothesis. Historical-comparative reconstruction is a science, but it also has elements of the artistic. The way an experienced comparativist can draw on a variety of disparate disciplines including phonology, grammar, psycholinguistics, sociolinguistics, dialectology, philology, history, anthropology, etc., then paint a portrait of a long-dead language is a beauty to see. I am afraid that this study will be a rather sterile exercise in comparison, i.e. much method and little art, given how impoverished is my knowledge of the history and prehistory of Lampung, and my lack of abilities in many of the related disciplines. Nevertheless by making my methodological decisions explicit one could hope that someone following afterward will at least have specifics to critique. 1.7 WordCorr Mechanical processing of the approximately 6500 Lampungic lexemes gathered was done with a PC program called WORDCORR. It allows the user to keep track of the huge amounts of data involved with historical comparative linguistics, gives a structure for the data to be entered uniformly, and allows the user to manipulate and organize data according to several different analytical viewpoints. Several different language projects can be worked on at once within WordCorr as well and passed to other linguists for review and comments. It is able to generate an exhaustive list of correspondence sets along with any conditioning environments a linguist wishes to posit, and reconstructions of all the words used for comparison. 2 I used WordCorr to keep track of correspondence sets, to help me find the regularly recurring changes from Proto-Lampungic to modern Lampungic isolects, and to reconstruct PLP phonemes and word forms. I will now briefly detail the steps I went through in processing the data. After the wordlists were gathered and entered in electronically, they were placed together in an Excel template, then processed with a program called PALMSURV CONVERTER (http://sourceforge.net/projects/psconverter) to get into a form which could be imported (data and metadata) into WordCorr. The next step was to decide which lists would be included in the analysis and in which order. This is called a VIEW; decisions made related to this are discussed in 2.1. One of the decisions which had to be made was the THRESHOLD percentage at which WordCorr would include correspondence sets for analysis. For example, in a collection of ten lists, for the gloss lightning there may be three different cognate sets, one with five representatives and the others with just a few. The default percentage in WordCorr is set at 50% which basically means only one correspondence set per gloss will be selected for analysis. Given the large number of wordlists in this collection and a desire to be as thorough as possible, I set the Threshold at 10%. At 10%, the threshold catches for analysis any word with at least 3 reflexes. This seemed appropriate given that my goal was not simply to reconstruct a proto-language but 2 For a more complete description of WordCorr and its abilities go to The user guide found there has complete information on this product.
12 12 also catch and analyze reflexes of higher-order reconstructions, primarily Proto-Malayo- Polynesian. This requires a larger data set. Once the cognates are set into groups, the next task was to ALIGN them so that, in a word like rua two, the reflexes of *r lined up with each other, *u, *a, etc. Once the segments are lined up, I assigned a protosegment to each correspondence set of each cognate group, and an environment, like word-final, vowel-medial, etc. These environments are detailed in 3. Once that was finished for all 350 glosses, I was able to look at all the correspondence sets together and REFINE my analysis, grouping or splitting environments or protosegments, noting any irregularities. For example, when I observed that reflexes of *ǝ in ultimate syllables behaved a certain way sometimes but differently in others, I was able to look at all the examples together and determine that a further conditioning environment needed to be specified to explain the extant reflexes. (This particular example is discussed in ) Finally, I was able to export my analyses, which appear, after a bit of massaging, in the data sections of this paper. 2 Lampung today 2.1 Data sources The primary data sources for this study, besides those mentioned in 1.4, are twenty-three wordlists, some of which are 350 items in length, others 200. See Table 1 for details, including the order in which the data in later sections will be presented. Map 2 shows much of the same information in visual form. Table 1 Wordlist sites Village Short name Dialect (Subdialect) Source Length 1. Kayu Agung Asli KAAsli KOMERING (Kayu Agung Asli) SIL Adumanis Kom-Adu KOMERING (Ulu) SIL Pulau Gemantung KomIlir KOMERING (Ilir) SIL Perjaya Kom-Jaya KOMERING (Ulu) SIL Damarpura Kom-Dpur KOMERING (Ulu) SIL Tihang Daya API (Daya) SIL Pilla Ranau API (Ranau) SIL Buay Nyerupa Sukau API (Sukau) SIL Banjar Agung Krui API (Krui) SIL Kota Besi Belalau API (Belalau) SIL Mesir Udik WayKanan API (Way Kanan) SIL Kandang Besi KotAgung API (Kota Agung) SIL Sukanegeri Jaya TalaPada API (Talang Padang) SIL Way Lima WayLima API (Way Lima) Walker Banjar Ketapang Sungkai API (Sungkai) SIL Negeri Kepayungan Pubian API (Pubian) SIL Tengkujuh Kalianda API (Kalianda) SIL Nibung Melintin NYO (Melinting) SIL Jabung Jabung API (Jabung) SIL Paku KAPend KOMERING (Kayu Agung Pendatang) SIL Nyampir Sukadana NYO (Abung/Sukadana) SIL Blambangan Pagar KotaBumi NYO (Abung/Kotabumi) SIL Ujung Gunung Menggala NYO (Menggala/Tulang Bawang) SIL 350
13 13 Map 2 LP wordlist sites One may notice that many of the SIL wordlist sites are also sites of lists from Walker (1975) and wonder why we revisited those sites. This following in Walker s footsteps, so to speak, was
14 14 deliberate, because Walker s lists (excepting Way Lima) were not elicited by him, a trained linguist, but were rather from native speakers from those areas but living elsewhere, written in orthographic script. So one could say that the later SIL lists from these areas are simply a refinement and check on the earlier work. 3 The exception to this pattern is Way Lima, where Walker himself had done extensive fieldwork, and we did not sample there. The SIL teams also sampled several areas from where Walker had not published lists. I am confident that the lists used in this study provide a fairly comprehensive representation of the multiplicity of LP isolects. 2.2 Phonology This section summarizes the phonology of the Lampungic speech varieties based upon our research. 4 Some selected differences with and between the two phonologies published by Walker (1976) and Abdurrahman and Yallop (1979) will also be discussed briefly. It should be noted that Walker s phonology was done on Way Lima, a village from the southern part of the region. Abdurrahman is from a village in the Komering area. The phonology presented here attempts to look at the whole cluster of Lampungic speech varieties at once, thus variations between this presentation and what is actually found at any given location will certainly differ to a small degree Consonants Table 2 displays the consonant phoneme inventory of Lampung. Table 2 Basic consonant phonemes of the Lampungic cluster Labial Alveolar Palatal Velar Glottal Stops /p/, /b/ /t/, /d/ /k/, /g/ /ʔ/ Fricatives /s/, (/z/ 5 ) /r/ 6 /h/ Affricates /c/, /j/ 7 Nasals /m/ /n/ /ɲ/ /ŋ/ Liquids /l/ Trill Semivowels /w/ /y/ Voiceless stops occur word initial, word medial, and word final. Word final stops are generally unreleased. Voiced stops generally do not occur word final. There seems to be only moderate evidence for a phonemic glottal stop. Final oral stops often undergo debuccalization (loss of oral place features). /r/ has a range of phonetic realizations but is most often a velar or uvular fricative [x], [ɣ], [χ], [ʁ]. There is minor disagreement between the two earlier phonologies about /r/, described as an apical trill by Abdurrahman and as a voiceless velar fricative by Walker. 3 The SIL teams did other work in these areas including sociolinguistic questionnaires, sociolinguistic observation, recorded text testing, etc. See Hanawalt et al. (in progress) for further details. 4 Note: this section is taken near-verbatim from Hanawalt et al. (in progress) which functions as a sort of umbrella report for other subsidiary papers such as this. 5 /z/ only occurs in loanwords. 6 /r/ here is used to refer to a range of different phonemic realizations of *r, depending upon the speech variety in question. 7 In this paper, [c] and [j] will be used to represent the IPA affricates [tʃ] and [dʒ] respectively.
15 15 Walker stated that this phoneme (written as /x/ in 1976 and as /r/ in his 1975 wordlists) occurs in all major environments and is sometimes voiced intervocalically. Walker noted that [r] (apical trill) occurs in unassimilated loanwords (1976:3) and alternates with [x] in many cases. The nasals occur word initial, word medial and word final, with the exception of /ɲ/, which does not occur word final. /l/ occurs word initial, word medial, and word final. /w/ and /y/ occur word-initially and word-medially and, depending on one s analysis, word-finally as part of diphthongs discussed below. Both phonemes occur word-medially in positions where they are not considered as transitions from [u] and [i] respectively. Gemination Gemination, particularly consonant gemination, is a prominent feature in Lampung. It is not easy to generalize about it except to say that it happens most frequently in Nyo, less so but still frequently in Api, and almost never (at least as we transcribed it) in Komering. Our informants often did not agree among themselves which ones were geminate but one can see that the phenomenon as we documented it clusters around specific lexemes. Several cases each of gemination are recorded for every consonant in medial only position (either between vowels or as part of a consonant cluster) except /ɲ/, /ŋ/, /s/, /w/ and /y/. Gemination is most frequently associated with one of two related environments: penultimate schwa reduction of voiceless nasal-stop clusters to the stop component. It therefore can give a clue to reconstructions. See the case of pǝkːul roof ( 4.1.3). Metathesis Metathesis seems to have been a fairly common process in Lampungic. From the correspondence sets can be counted at least twenty lexemes in which metathesis occurred in one or more of the Lampungic isolects. It most frequently occurred with consonants, e.g. *rihuʔ cloud hiruʔ and *gǝlar name gǝral, but also with vowels, often with some fairly complex transformations, e.g. *siwa nine suay and lahya ginger liha. A few instances of metathesis can be attributed to PLP; see 5.4. Debuccalization A final phonological change happening on a irregular but very frequent basis is DEBUCCALIZATION the process in which an oral consonant, in LP s case a final voiceless stop, becomes a glottal stop. Debuccalization is seen to occur in all varieties of Lampungic except KAPend. Table 3 illustrates this phenomenon. This change is attested in the cases of *p > ʔ and *t > ʔ. It is also seen in the case of *k > ʔ, but the attestations are less consistent. Table 3 Examples of debuccalization gloss PLP KAPend KAAsli KotaBumi bitter *pahit pahit pahiʔ pahiʔ smoke *hasəp hasop hasoʔ asǝʔ needle *sǝrǝp sowop soʁoʔ sǝʁǝʔ husk of rice *huǝt huot huoʔ uǝʔ
16 Vowels Table 4 displays the vowel and diphthong phonemes found in the Lampungic speech varieties. Table 4 Basic vowel and diphthong phonemes of the Lampungic cluster Front Central Back unrounded unrounded rounded Close /i/ /u/ Mid (/e/) /ə/ (/o/) Open /a/ Diphthongs /ay/ /aw/ /uy/ Abdurrahman posits the phoneme /o/ for Komering. However, we submit that most if not all occurrences of [o] can be more accurately analyzed as allophones of /ə/. See for more information. Walker posits the phoneme /e/ in addition to /i/ and /ə/ for Way Lima. Our preference, however, is to reanalyze most occurrences of [e] as allophones of /i/. Walker apparently did not preserve the distinction between [ə] and [e] in his wordlists, as both phones are written using e. In addition, the examples he gives in his phonology for /e/ are likely all borrowed words. It seems that /ǝ/ cannot occur before nasal-stop clusters. Vowel sequences do occur, but a syllable break always occurs between them in our data. Such sequences are distinguished from the diphthongs /aw/, /ay/, and /uy/. A more comprehensive phonology of the cluster or of its individual speech varieties is beyond the scope of this paper. However, the reader is referred to several works cited in Hanawalt et al. 2.3 Dialect differences in detail In a work such as this it is crucial to understand the processes which have led to the consistent differences between the daughter isolects of the proto-language. By analyzing each cognate word for each lexical item across the speech varieties, it is possible to view the similarity or difference in individual sounds segments across the varieties. By looking at multiple similarities and differences in this way, patterns of phonological retentions or changes may emerge across these word correspondences. Having a grasp of these patterns allows us to make more accurate reconstructions and bracket certain words as loans Realization of *ə The form that ultimate *ə takes in many of the speech varieties provides a measure on which to make a large grouping. This realization of *ə in the ultimate syllable is demonstrated in Table For the purpose of clarity, some details of phonetic transcription have been omitted from the data presented in these and following examples.
17 17 Table 5 Examples of ultimate *ə > [o] gloss PLP Melintin KAAsli Krui earth *tanəh tanəh tanoh tanoh itch *gatəl gatəl gatol gatol sit *məjəŋ məjːəŋ mojoŋ məjoŋ suck *hisəp isəp hisoʔ ŋisop *ə > o nearly consistently in the ultimate syllable in the varieties shown in Map 3 below, forming a chain from Kayu Agung in the north curving westward and ending at Kalianda in the south. This chain includes the areas in Walker s PESISIR group, plus the area Jabung. Our data show that three Nyo varieties exhibit this change only in specific environments. In Menggala this change follows a clear pattern. *ə when followed by oral consonants is reflected as ə; before non-oral consonants (/h/ and /ʔ/), *ə appears as o. Sukadana shows the same split as MGL, but in non-oral environments the reflex is a rather than Menggala s o. Kotabumi (in Walker s list) consistently shows a split in environment and reflexes identical to Sukadana. In Melintin there is no split; all *ə are reflected as ə. The first two examples in Table 6 show the *ə reflex in environments before non-oral consonants; the third and fourth examples are in an environment before oral consonants. Table 6 Examples of split ultimate *ə realization in Nyo areas gloss PLP Melintin Sukadana Kota Bumi Menggala (Walker) smoke *hasəp asəʔ asaʔ asaʔ asoʔ near *parə(ʔ) paʁəʔ paʁaʔ paʁaʔ paʁoʔ black *harəŋ aʁəŋ aʁəŋ aʁəŋ aʁəŋ sea *lawət lawət lawət lawət lawət Interestingly, the Blambangan Pagar (SIL s KotaBumi) wordlist taken in 2005, which is approximately ten kilometers from Kota Bumi (and was considered to be the center of the same speech variety by the locals) shows an even more complex split. As with the four isolects above, basically all of the oral environments have [ə]. But there is an additional split WITHIN the non-oral environment conditioned on the penultimate vowel; one could call it a dissimilation rule: if the penultimate vowel is ə, the ultimate vowel will be reflected as [a]. With other vowels in the penult (i, u or a), generally [ə] will appear. In phonological notation, the pattern in Kota Bumi is something like this: /ǝ/ -> [ǝ]/_c [+oral] # -> [ǝ]/v [-ə] (C)_C [-oral] # -> [a]/v [+ə] (C)_C [-oral] # The examples in Table 7 demonstrate these seemingly variant realizations. Table 7 Examples of ultimate *ə in non-oral environments in Blambangan Pagar (KotaBumi) gloss PLP Melintin Sukadana Blambangan Pagar (KotaBumi) smoke *hasəp asəʔ asaʔ asəʔ earth *tanǝh tanǝh tanah tanǝh husk of rice *huǝt - uaʔ uǝʔ cut/hack *pǝlǝʔ pǝlǝʔ - pǝlaʔ hungry *bǝtǝh - bǝtah bǝtah
18 18 Many varieties also display a uniformly different realization of *ə as o in the PENULTIMATE syllable, as demonstrated in Table 8. Table 8 Examples of penultimate *ə > [o] gloss PLP Krui KAPend KAAsli Kom-Adu sand *hənay həni honi honi honi pestle *həlu həlu holu holu holu worm *gələŋ gəloŋ goloŋ goloŋ goloŋ g sugar cane *təbu təbu tobu tobu tobu As shown in Map 3, this change occurs in the varieties along the Komering River, from Adumanis downstream through both Kayu Agung Asli ( Native Kayu Agung ) and Kayu Agung Pendatang ( Newcomer Kayu Agung ; Paku village).
19 Map 3 Penultimate and ultimate *ə 19
20 Vowel lowering *u is frequently lowered in ultimate closed (V1) position in KAAsli. Both high vowels are frequently lowered in penultimate and ultimate closed syllables in Menggala Nasal consonant cluster reduction Reflexes of consonant clusters are usually what one would expect (i.e. a nasal followed by a homorganic stop, both of relatively equal prominence) but there are principled exceptions. If the cluster includes a VOICED stop, e.g. *imbun fog, if one of the members is to be elided it will be the stop. See Table 9. Table 9 Examples of *NS > N with voiced stops gloss PLP WayKanan Jabung Melintin KAPend white *handaʔ handaʔ an d aʔ n d aʔ han d aʔ mother *induʔ induʔ nuʔ nːuʔ on d oʔ gloss PLP KomIlir WayKanan KAGA to boil (water) *ruŋgaʔ ʁuŋgaʔ ʁuŋgaʔ ʁuŋaʔ In all other cases of elision (clusters with voiceless stops, voiced and voiceless affricates, liquids), it is the nasal which is lost, often with neutralization of the preceding vowel to schwa and/or a geminate stop. E.g. *punti banana in some varieties is reflected as pǝtːi. This is substantially more common than elision of voiced stops. This consistent pattern has led me to realize that certain correspondence sets were unlikely. For example, I had earlier grouped timbuʔ and cibuk dipper but it seems prudent to separate them because the stop is voiced Deletion of initial *h There are two main clusters of a few speech varieties each which demonstrate loss of *h at the beginning of a word, as in Table 10. Table 10 *h deletion word initial gloss PLP KomIlir Krui Menggala head PMP *qulu > PLP *hulu hulu ulu ulew smoke PMP *qasep > PLP *hasəp hasoʔ asoʔ asoʔ The first group of speech varieties loses the initial h in almost every instance. This first group includes the Nyo group plus Jabung. The second group of speech varieties that elides *h only does so part of the time but in the same words. This group consists of Ranau, Sukau and Krui. As Map 4 displays, the members of each of those groups share close geographic proximity. Sporadic loss of initial *h occurs in some other varieties as well.